Feature Selection and Ranking via Simultaneous Perturbation Stochastic Approximation R package spFSR version 2.0.4 An implementation of feature selection, weighting and ranking via simultaneous perturbation stochastic approximation (SPSA). The SPSA-FSR algorithm searches for a locally optimal set of features that yield the best predictive performance using some error measures such as mean squared error (for regression problems) and accuracy rate (for classification problems). Software Depends: mlr3 (>= 0.14.0), future (>= 1.28.0), tictoc (>= 1.0) Imports: mlr3pipelines (>= 0.4.2), mlr3learners (>= 0.5.4), ranger (>= 0.14.1), parallel (>= 3.4.2), ggplot2 (>= 2.2.1), lgr (>= 0.4.4) Suggests: caret (>= 6.0), MASS (>= 7.3) David Akman <david.v.akman@gmail.com> Comprehensive R Archive Network (CRAN) David Akman [aut, cre], Babak Abbasi [aut, ctb], Yong Kai Wong [aut, ctb], Guo Feng Anders Yeo [aut, ctb], Zeren D. Yenice [ctb] GPL-3 2023-03-17 application/tgz https://CRAN.R-project.org/package=spFSR doi:10.32614/CRAN.package.spFSR